1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the AF_INET socket handler. 7 * 8 * Version: @(#)sock.h 1.0.4 05/13/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Florian La Roche <flla@stud.uni-sb.de> 14 * 15 * Fixes: 16 * Alan Cox : Volatiles in skbuff pointers. See 17 * skbuff comments. May be overdone, 18 * better to prove they can be removed 19 * than the reverse. 20 * Alan Cox : Added a zapped field for tcp to note 21 * a socket is reset and must stay shut up 22 * Alan Cox : New fields for options 23 * Pauline Middelink : identd support 24 * Alan Cox : Eliminate low level recv/recvfrom 25 * David S. Miller : New socket lookup architecture. 26 * Steve Whitehouse: Default routines for sock_ops 27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made 28 * protinfo be just a void pointer, as the 29 * protocol specific parts were moved to 30 * respective headers and ipv4/v6, etc now 31 * use private slabcaches for its socks 32 * Pedro Hortas : New flags field for socket options 33 * 34 * 35 * This program is free software; you can redistribute it and/or 36 * modify it under the terms of the GNU General Public License 37 * as published by the Free Software Foundation; either version 38 * 2 of the License, or (at your option) any later version. 39 */ 40 #ifndef _SOCK_H 41 #define _SOCK_H 42 43 #include <linux/kernel.h> 44 #include <linux/list.h> 45 #include <linux/timer.h> 46 #include <linux/cache.h> 47 #include <linux/module.h> 48 #include <linux/lockdep.h> 49 #include <linux/netdevice.h> 50 #include <linux/skbuff.h> /* struct sk_buff */ 51 #include <linux/mm.h> 52 #include <linux/security.h> 53 54 #include <linux/filter.h> 55 56 #include <asm/atomic.h> 57 #include <net/dst.h> 58 #include <net/checksum.h> 59 60 /* 61 * This structure really needs to be cleaned up. 62 * Most of it is for TCP, and not used by any of 63 * the other protocols. 64 */ 65 66 /* Define this to get the SOCK_DBG debugging facility. */ 67 #define SOCK_DEBUGGING 68 #ifdef SOCK_DEBUGGING 69 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ 70 printk(KERN_DEBUG msg); } while (0) 71 #else 72 /* Validate arguments and do nothing */ 73 static void inline int __attribute__ ((format (printf, 2, 3))) 74 SOCK_DEBUG(struct sock *sk, const char *msg, ...) 75 { 76 } 77 #endif 78 79 /* This is the per-socket lock. The spinlock provides a synchronization 80 * between user contexts and software interrupt processing, whereas the 81 * mini-semaphore synchronizes multiple users amongst themselves. 82 */ 83 typedef struct { 84 spinlock_t slock; 85 int owned; 86 wait_queue_head_t wq; 87 /* 88 * We express the mutex-alike socket_lock semantics 89 * to the lock validator by explicitly managing 90 * the slock as a lock variant (in addition to 91 * the slock itself): 92 */ 93 #ifdef CONFIG_DEBUG_LOCK_ALLOC 94 struct lockdep_map dep_map; 95 #endif 96 } socket_lock_t; 97 98 struct sock; 99 struct proto; 100 struct net; 101 102 /** 103 * struct sock_common - minimal network layer representation of sockets 104 * @skc_family: network address family 105 * @skc_state: Connection state 106 * @skc_reuse: %SO_REUSEADDR setting 107 * @skc_bound_dev_if: bound device index if != 0 108 * @skc_node: main hash linkage for various protocol lookup tables 109 * @skc_bind_node: bind hash linkage for various protocol lookup tables 110 * @skc_refcnt: reference count 111 * @skc_hash: hash value used with various protocol lookup tables 112 * @skc_prot: protocol handlers inside a network family 113 * @skc_net: reference to the network namespace of this socket 114 * 115 * This is the minimal network layer representation of sockets, the header 116 * for struct sock and struct inet_timewait_sock. 117 */ 118 struct sock_common { 119 unsigned short skc_family; 120 volatile unsigned char skc_state; 121 unsigned char skc_reuse; 122 int skc_bound_dev_if; 123 struct hlist_node skc_node; 124 struct hlist_node skc_bind_node; 125 atomic_t skc_refcnt; 126 unsigned int skc_hash; 127 struct proto *skc_prot; 128 #ifdef CONFIG_NET_NS 129 struct net *skc_net; 130 #endif 131 }; 132 133 /** 134 * struct sock - network layer representation of sockets 135 * @__sk_common: shared layout with inet_timewait_sock 136 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN 137 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings 138 * @sk_lock: synchronizer 139 * @sk_rcvbuf: size of receive buffer in bytes 140 * @sk_sleep: sock wait queue 141 * @sk_dst_cache: destination cache 142 * @sk_dst_lock: destination cache lock 143 * @sk_policy: flow policy 144 * @sk_rmem_alloc: receive queue bytes committed 145 * @sk_receive_queue: incoming packets 146 * @sk_wmem_alloc: transmit queue bytes committed 147 * @sk_write_queue: Packet sending queue 148 * @sk_async_wait_queue: DMA copied packets 149 * @sk_omem_alloc: "o" is "option" or "other" 150 * @sk_wmem_queued: persistent queue size 151 * @sk_forward_alloc: space allocated forward 152 * @sk_allocation: allocation mode 153 * @sk_sndbuf: size of send buffer in bytes 154 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, 155 * %SO_OOBINLINE settings 156 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets 157 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) 158 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) 159 * @sk_gso_max_size: Maximum GSO segment size to build 160 * @sk_lingertime: %SO_LINGER l_linger setting 161 * @sk_backlog: always used with the per-socket spinlock held 162 * @sk_callback_lock: used with the callbacks in the end of this struct 163 * @sk_error_queue: rarely used 164 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, 165 * IPV6_ADDRFORM for instance) 166 * @sk_err: last error 167 * @sk_err_soft: errors that don't cause failure but are the cause of a 168 * persistent failure not just 'timed out' 169 * @sk_drops: raw/udp drops counter 170 * @sk_ack_backlog: current listen backlog 171 * @sk_max_ack_backlog: listen backlog set in listen() 172 * @sk_priority: %SO_PRIORITY setting 173 * @sk_type: socket type (%SOCK_STREAM, etc) 174 * @sk_protocol: which protocol this socket belongs in this network family 175 * @sk_peercred: %SO_PEERCRED setting 176 * @sk_rcvlowat: %SO_RCVLOWAT setting 177 * @sk_rcvtimeo: %SO_RCVTIMEO setting 178 * @sk_sndtimeo: %SO_SNDTIMEO setting 179 * @sk_filter: socket filtering instructions 180 * @sk_protinfo: private area, net family specific, when not using slab 181 * @sk_timer: sock cleanup timer 182 * @sk_stamp: time stamp of last packet received 183 * @sk_socket: Identd and reporting IO signals 184 * @sk_user_data: RPC layer private data 185 * @sk_sndmsg_page: cached page for sendmsg 186 * @sk_sndmsg_off: cached offset for sendmsg 187 * @sk_send_head: front of stuff to transmit 188 * @sk_security: used by security modules 189 * @sk_mark: generic packet mark 190 * @sk_write_pending: a write to stream socket waits to start 191 * @sk_state_change: callback to indicate change in the state of the sock 192 * @sk_data_ready: callback to indicate there is data to be processed 193 * @sk_write_space: callback to indicate there is bf sending space available 194 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) 195 * @sk_backlog_rcv: callback to process the backlog 196 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 197 */ 198 struct sock { 199 /* 200 * Now struct inet_timewait_sock also uses sock_common, so please just 201 * don't add nothing before this first member (__sk_common) --acme 202 */ 203 struct sock_common __sk_common; 204 #define sk_family __sk_common.skc_family 205 #define sk_state __sk_common.skc_state 206 #define sk_reuse __sk_common.skc_reuse 207 #define sk_bound_dev_if __sk_common.skc_bound_dev_if 208 #define sk_node __sk_common.skc_node 209 #define sk_bind_node __sk_common.skc_bind_node 210 #define sk_refcnt __sk_common.skc_refcnt 211 #define sk_hash __sk_common.skc_hash 212 #define sk_prot __sk_common.skc_prot 213 #define sk_net __sk_common.skc_net 214 unsigned char sk_shutdown : 2, 215 sk_no_check : 2, 216 sk_userlocks : 4; 217 unsigned char sk_protocol; 218 unsigned short sk_type; 219 int sk_rcvbuf; 220 socket_lock_t sk_lock; 221 /* 222 * The backlog queue is special, it is always used with 223 * the per-socket spinlock held and requires low latency 224 * access. Therefore we special case it's implementation. 225 */ 226 struct { 227 struct sk_buff *head; 228 struct sk_buff *tail; 229 } sk_backlog; 230 wait_queue_head_t *sk_sleep; 231 struct dst_entry *sk_dst_cache; 232 struct xfrm_policy *sk_policy[2]; 233 rwlock_t sk_dst_lock; 234 atomic_t sk_rmem_alloc; 235 atomic_t sk_wmem_alloc; 236 atomic_t sk_omem_alloc; 237 int sk_sndbuf; 238 struct sk_buff_head sk_receive_queue; 239 struct sk_buff_head sk_write_queue; 240 struct sk_buff_head sk_async_wait_queue; 241 int sk_wmem_queued; 242 int sk_forward_alloc; 243 gfp_t sk_allocation; 244 int sk_route_caps; 245 int sk_gso_type; 246 unsigned int sk_gso_max_size; 247 int sk_rcvlowat; 248 unsigned long sk_flags; 249 unsigned long sk_lingertime; 250 struct sk_buff_head sk_error_queue; 251 struct proto *sk_prot_creator; 252 rwlock_t sk_callback_lock; 253 int sk_err, 254 sk_err_soft; 255 atomic_t sk_drops; 256 unsigned short sk_ack_backlog; 257 unsigned short sk_max_ack_backlog; 258 __u32 sk_priority; 259 struct ucred sk_peercred; 260 long sk_rcvtimeo; 261 long sk_sndtimeo; 262 struct sk_filter *sk_filter; 263 void *sk_protinfo; 264 struct timer_list sk_timer; 265 ktime_t sk_stamp; 266 struct socket *sk_socket; 267 void *sk_user_data; 268 struct page *sk_sndmsg_page; 269 struct sk_buff *sk_send_head; 270 __u32 sk_sndmsg_off; 271 int sk_write_pending; 272 void *sk_security; 273 __u32 sk_mark; 274 /* XXX 4 bytes hole on 64 bit */ 275 void (*sk_state_change)(struct sock *sk); 276 void (*sk_data_ready)(struct sock *sk, int bytes); 277 void (*sk_write_space)(struct sock *sk); 278 void (*sk_error_report)(struct sock *sk); 279 int (*sk_backlog_rcv)(struct sock *sk, 280 struct sk_buff *skb); 281 void (*sk_destruct)(struct sock *sk); 282 }; 283 284 /* 285 * Hashed lists helper routines 286 */ 287 static inline struct sock *__sk_head(const struct hlist_head *head) 288 { 289 return hlist_entry(head->first, struct sock, sk_node); 290 } 291 292 static inline struct sock *sk_head(const struct hlist_head *head) 293 { 294 return hlist_empty(head) ? NULL : __sk_head(head); 295 } 296 297 static inline struct sock *sk_next(const struct sock *sk) 298 { 299 return sk->sk_node.next ? 300 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; 301 } 302 303 static inline int sk_unhashed(const struct sock *sk) 304 { 305 return hlist_unhashed(&sk->sk_node); 306 } 307 308 static inline int sk_hashed(const struct sock *sk) 309 { 310 return !sk_unhashed(sk); 311 } 312 313 static __inline__ void sk_node_init(struct hlist_node *node) 314 { 315 node->pprev = NULL; 316 } 317 318 static __inline__ void __sk_del_node(struct sock *sk) 319 { 320 __hlist_del(&sk->sk_node); 321 } 322 323 static __inline__ int __sk_del_node_init(struct sock *sk) 324 { 325 if (sk_hashed(sk)) { 326 __sk_del_node(sk); 327 sk_node_init(&sk->sk_node); 328 return 1; 329 } 330 return 0; 331 } 332 333 /* Grab socket reference count. This operation is valid only 334 when sk is ALREADY grabbed f.e. it is found in hash table 335 or a list and the lookup is made under lock preventing hash table 336 modifications. 337 */ 338 339 static inline void sock_hold(struct sock *sk) 340 { 341 atomic_inc(&sk->sk_refcnt); 342 } 343 344 /* Ungrab socket in the context, which assumes that socket refcnt 345 cannot hit zero, f.e. it is true in context of any socketcall. 346 */ 347 static inline void __sock_put(struct sock *sk) 348 { 349 atomic_dec(&sk->sk_refcnt); 350 } 351 352 static __inline__ int sk_del_node_init(struct sock *sk) 353 { 354 int rc = __sk_del_node_init(sk); 355 356 if (rc) { 357 /* paranoid for a while -acme */ 358 WARN_ON(atomic_read(&sk->sk_refcnt) == 1); 359 __sock_put(sk); 360 } 361 return rc; 362 } 363 364 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) 365 { 366 hlist_add_head(&sk->sk_node, list); 367 } 368 369 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) 370 { 371 sock_hold(sk); 372 __sk_add_node(sk, list); 373 } 374 375 static __inline__ void __sk_del_bind_node(struct sock *sk) 376 { 377 __hlist_del(&sk->sk_bind_node); 378 } 379 380 static __inline__ void sk_add_bind_node(struct sock *sk, 381 struct hlist_head *list) 382 { 383 hlist_add_head(&sk->sk_bind_node, list); 384 } 385 386 #define sk_for_each(__sk, node, list) \ 387 hlist_for_each_entry(__sk, node, list, sk_node) 388 #define sk_for_each_from(__sk, node) \ 389 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ 390 hlist_for_each_entry_from(__sk, node, sk_node) 391 #define sk_for_each_continue(__sk, node) \ 392 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ 393 hlist_for_each_entry_continue(__sk, node, sk_node) 394 #define sk_for_each_safe(__sk, node, tmp, list) \ 395 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) 396 #define sk_for_each_bound(__sk, node, list) \ 397 hlist_for_each_entry(__sk, node, list, sk_bind_node) 398 399 /* Sock flags */ 400 enum sock_flags { 401 SOCK_DEAD, 402 SOCK_DONE, 403 SOCK_URGINLINE, 404 SOCK_KEEPOPEN, 405 SOCK_LINGER, 406 SOCK_DESTROY, 407 SOCK_BROADCAST, 408 SOCK_TIMESTAMP, 409 SOCK_ZAPPED, 410 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ 411 SOCK_DBG, /* %SO_DEBUG setting */ 412 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ 413 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ 414 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ 415 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ 416 }; 417 418 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) 419 { 420 nsk->sk_flags = osk->sk_flags; 421 } 422 423 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) 424 { 425 __set_bit(flag, &sk->sk_flags); 426 } 427 428 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) 429 { 430 __clear_bit(flag, &sk->sk_flags); 431 } 432 433 static inline int sock_flag(struct sock *sk, enum sock_flags flag) 434 { 435 return test_bit(flag, &sk->sk_flags); 436 } 437 438 static inline void sk_acceptq_removed(struct sock *sk) 439 { 440 sk->sk_ack_backlog--; 441 } 442 443 static inline void sk_acceptq_added(struct sock *sk) 444 { 445 sk->sk_ack_backlog++; 446 } 447 448 static inline int sk_acceptq_is_full(struct sock *sk) 449 { 450 return sk->sk_ack_backlog > sk->sk_max_ack_backlog; 451 } 452 453 /* 454 * Compute minimal free write space needed to queue new packets. 455 */ 456 static inline int sk_stream_min_wspace(struct sock *sk) 457 { 458 return sk->sk_wmem_queued >> 1; 459 } 460 461 static inline int sk_stream_wspace(struct sock *sk) 462 { 463 return sk->sk_sndbuf - sk->sk_wmem_queued; 464 } 465 466 extern void sk_stream_write_space(struct sock *sk); 467 468 static inline int sk_stream_memory_free(struct sock *sk) 469 { 470 return sk->sk_wmem_queued < sk->sk_sndbuf; 471 } 472 473 /* The per-socket spinlock must be held here. */ 474 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb) 475 { 476 if (!sk->sk_backlog.tail) { 477 sk->sk_backlog.head = sk->sk_backlog.tail = skb; 478 } else { 479 sk->sk_backlog.tail->next = skb; 480 sk->sk_backlog.tail = skb; 481 } 482 skb->next = NULL; 483 } 484 485 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 486 { 487 return sk->sk_backlog_rcv(sk, skb); 488 } 489 490 #define sk_wait_event(__sk, __timeo, __condition) \ 491 ({ int __rc; \ 492 release_sock(__sk); \ 493 __rc = __condition; \ 494 if (!__rc) { \ 495 *(__timeo) = schedule_timeout(*(__timeo)); \ 496 } \ 497 lock_sock(__sk); \ 498 __rc = __condition; \ 499 __rc; \ 500 }) 501 502 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); 503 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); 504 extern void sk_stream_wait_close(struct sock *sk, long timeo_p); 505 extern int sk_stream_error(struct sock *sk, int flags, int err); 506 extern void sk_stream_kill_queues(struct sock *sk); 507 508 extern int sk_wait_data(struct sock *sk, long *timeo); 509 510 struct request_sock_ops; 511 struct timewait_sock_ops; 512 struct inet_hashinfo; 513 struct raw_hashinfo; 514 515 /* Networking protocol blocks we attach to sockets. 516 * socket layer -> transport layer interface 517 * transport -> network interface is defined by struct inet_proto 518 */ 519 struct proto { 520 void (*close)(struct sock *sk, 521 long timeout); 522 int (*connect)(struct sock *sk, 523 struct sockaddr *uaddr, 524 int addr_len); 525 int (*disconnect)(struct sock *sk, int flags); 526 527 struct sock * (*accept) (struct sock *sk, int flags, int *err); 528 529 int (*ioctl)(struct sock *sk, int cmd, 530 unsigned long arg); 531 int (*init)(struct sock *sk); 532 void (*destroy)(struct sock *sk); 533 void (*shutdown)(struct sock *sk, int how); 534 int (*setsockopt)(struct sock *sk, int level, 535 int optname, char __user *optval, 536 int optlen); 537 int (*getsockopt)(struct sock *sk, int level, 538 int optname, char __user *optval, 539 int __user *option); 540 #ifdef CONFIG_COMPAT 541 int (*compat_setsockopt)(struct sock *sk, 542 int level, 543 int optname, char __user *optval, 544 int optlen); 545 int (*compat_getsockopt)(struct sock *sk, 546 int level, 547 int optname, char __user *optval, 548 int __user *option); 549 #endif 550 int (*sendmsg)(struct kiocb *iocb, struct sock *sk, 551 struct msghdr *msg, size_t len); 552 int (*recvmsg)(struct kiocb *iocb, struct sock *sk, 553 struct msghdr *msg, 554 size_t len, int noblock, int flags, 555 int *addr_len); 556 int (*sendpage)(struct sock *sk, struct page *page, 557 int offset, size_t size, int flags); 558 int (*bind)(struct sock *sk, 559 struct sockaddr *uaddr, int addr_len); 560 561 int (*backlog_rcv) (struct sock *sk, 562 struct sk_buff *skb); 563 564 /* Keeping track of sk's, looking them up, and port selection methods. */ 565 void (*hash)(struct sock *sk); 566 void (*unhash)(struct sock *sk); 567 int (*get_port)(struct sock *sk, unsigned short snum); 568 569 /* Keeping track of sockets in use */ 570 #ifdef CONFIG_PROC_FS 571 unsigned int inuse_idx; 572 #endif 573 574 /* Memory pressure */ 575 void (*enter_memory_pressure)(struct sock *sk); 576 atomic_t *memory_allocated; /* Current allocated memory. */ 577 atomic_t *sockets_allocated; /* Current number of sockets. */ 578 /* 579 * Pressure flag: try to collapse. 580 * Technical note: it is used by multiple contexts non atomically. 581 * All the __sk_mem_schedule() is of this nature: accounting 582 * is strict, actions are advisory and have some latency. 583 */ 584 int *memory_pressure; 585 int *sysctl_mem; 586 int *sysctl_wmem; 587 int *sysctl_rmem; 588 int max_header; 589 590 struct kmem_cache *slab; 591 unsigned int obj_size; 592 593 atomic_t *orphan_count; 594 595 struct request_sock_ops *rsk_prot; 596 struct timewait_sock_ops *twsk_prot; 597 598 union { 599 struct inet_hashinfo *hashinfo; 600 struct hlist_head *udp_hash; 601 struct raw_hashinfo *raw_hash; 602 } h; 603 604 struct module *owner; 605 606 char name[32]; 607 608 struct list_head node; 609 #ifdef SOCK_REFCNT_DEBUG 610 atomic_t socks; 611 #endif 612 }; 613 614 extern int proto_register(struct proto *prot, int alloc_slab); 615 extern void proto_unregister(struct proto *prot); 616 617 #ifdef SOCK_REFCNT_DEBUG 618 static inline void sk_refcnt_debug_inc(struct sock *sk) 619 { 620 atomic_inc(&sk->sk_prot->socks); 621 } 622 623 static inline void sk_refcnt_debug_dec(struct sock *sk) 624 { 625 atomic_dec(&sk->sk_prot->socks); 626 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", 627 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); 628 } 629 630 static inline void sk_refcnt_debug_release(const struct sock *sk) 631 { 632 if (atomic_read(&sk->sk_refcnt) != 1) 633 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", 634 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); 635 } 636 #else /* SOCK_REFCNT_DEBUG */ 637 #define sk_refcnt_debug_inc(sk) do { } while (0) 638 #define sk_refcnt_debug_dec(sk) do { } while (0) 639 #define sk_refcnt_debug_release(sk) do { } while (0) 640 #endif /* SOCK_REFCNT_DEBUG */ 641 642 643 #ifdef CONFIG_PROC_FS 644 /* Called with local bh disabled */ 645 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); 646 extern int sock_prot_inuse_get(struct net *net, struct proto *proto); 647 #else 648 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot, 649 int inc) 650 { 651 } 652 #endif 653 654 655 /* With per-bucket locks this operation is not-atomic, so that 656 * this version is not worse. 657 */ 658 static inline void __sk_prot_rehash(struct sock *sk) 659 { 660 sk->sk_prot->unhash(sk); 661 sk->sk_prot->hash(sk); 662 } 663 664 /* About 10 seconds */ 665 #define SOCK_DESTROY_TIME (10*HZ) 666 667 /* Sockets 0-1023 can't be bound to unless you are superuser */ 668 #define PROT_SOCK 1024 669 670 #define SHUTDOWN_MASK 3 671 #define RCV_SHUTDOWN 1 672 #define SEND_SHUTDOWN 2 673 674 #define SOCK_SNDBUF_LOCK 1 675 #define SOCK_RCVBUF_LOCK 2 676 #define SOCK_BINDADDR_LOCK 4 677 #define SOCK_BINDPORT_LOCK 8 678 679 /* sock_iocb: used to kick off async processing of socket ios */ 680 struct sock_iocb { 681 struct list_head list; 682 683 int flags; 684 int size; 685 struct socket *sock; 686 struct sock *sk; 687 struct scm_cookie *scm; 688 struct msghdr *msg, async_msg; 689 struct kiocb *kiocb; 690 }; 691 692 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) 693 { 694 return (struct sock_iocb *)iocb->private; 695 } 696 697 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) 698 { 699 return si->kiocb; 700 } 701 702 struct socket_alloc { 703 struct socket socket; 704 struct inode vfs_inode; 705 }; 706 707 static inline struct socket *SOCKET_I(struct inode *inode) 708 { 709 return &container_of(inode, struct socket_alloc, vfs_inode)->socket; 710 } 711 712 static inline struct inode *SOCK_INODE(struct socket *socket) 713 { 714 return &container_of(socket, struct socket_alloc, socket)->vfs_inode; 715 } 716 717 /* 718 * Functions for memory accounting 719 */ 720 extern int __sk_mem_schedule(struct sock *sk, int size, int kind); 721 extern void __sk_mem_reclaim(struct sock *sk); 722 723 #define SK_MEM_QUANTUM ((int)PAGE_SIZE) 724 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) 725 #define SK_MEM_SEND 0 726 #define SK_MEM_RECV 1 727 728 static inline int sk_mem_pages(int amt) 729 { 730 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; 731 } 732 733 static inline int sk_has_account(struct sock *sk) 734 { 735 /* return true if protocol supports memory accounting */ 736 return !!sk->sk_prot->memory_allocated; 737 } 738 739 static inline int sk_wmem_schedule(struct sock *sk, int size) 740 { 741 if (!sk_has_account(sk)) 742 return 1; 743 return size <= sk->sk_forward_alloc || 744 __sk_mem_schedule(sk, size, SK_MEM_SEND); 745 } 746 747 static inline int sk_rmem_schedule(struct sock *sk, int size) 748 { 749 if (!sk_has_account(sk)) 750 return 1; 751 return size <= sk->sk_forward_alloc || 752 __sk_mem_schedule(sk, size, SK_MEM_RECV); 753 } 754 755 static inline void sk_mem_reclaim(struct sock *sk) 756 { 757 if (!sk_has_account(sk)) 758 return; 759 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) 760 __sk_mem_reclaim(sk); 761 } 762 763 static inline void sk_mem_reclaim_partial(struct sock *sk) 764 { 765 if (!sk_has_account(sk)) 766 return; 767 if (sk->sk_forward_alloc > SK_MEM_QUANTUM) 768 __sk_mem_reclaim(sk); 769 } 770 771 static inline void sk_mem_charge(struct sock *sk, int size) 772 { 773 if (!sk_has_account(sk)) 774 return; 775 sk->sk_forward_alloc -= size; 776 } 777 778 static inline void sk_mem_uncharge(struct sock *sk, int size) 779 { 780 if (!sk_has_account(sk)) 781 return; 782 sk->sk_forward_alloc += size; 783 } 784 785 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) 786 { 787 skb_truesize_check(skb); 788 sock_set_flag(sk, SOCK_QUEUE_SHRUNK); 789 sk->sk_wmem_queued -= skb->truesize; 790 sk_mem_uncharge(sk, skb->truesize); 791 __kfree_skb(skb); 792 } 793 794 /* Used by processes to "lock" a socket state, so that 795 * interrupts and bottom half handlers won't change it 796 * from under us. It essentially blocks any incoming 797 * packets, so that we won't get any new data or any 798 * packets that change the state of the socket. 799 * 800 * While locked, BH processing will add new packets to 801 * the backlog queue. This queue is processed by the 802 * owner of the socket lock right before it is released. 803 * 804 * Since ~2.3.5 it is also exclusive sleep lock serializing 805 * accesses from user process context. 806 */ 807 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) 808 809 /* 810 * Macro so as to not evaluate some arguments when 811 * lockdep is not enabled. 812 * 813 * Mark both the sk_lock and the sk_lock.slock as a 814 * per-address-family lock class. 815 */ 816 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ 817 do { \ 818 sk->sk_lock.owned = 0; \ 819 init_waitqueue_head(&sk->sk_lock.wq); \ 820 spin_lock_init(&(sk)->sk_lock.slock); \ 821 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ 822 sizeof((sk)->sk_lock)); \ 823 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ 824 (skey), (sname)); \ 825 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ 826 } while (0) 827 828 extern void lock_sock_nested(struct sock *sk, int subclass); 829 830 static inline void lock_sock(struct sock *sk) 831 { 832 lock_sock_nested(sk, 0); 833 } 834 835 extern void release_sock(struct sock *sk); 836 837 /* BH context may only use the following locking interface. */ 838 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) 839 #define bh_lock_sock_nested(__sk) \ 840 spin_lock_nested(&((__sk)->sk_lock.slock), \ 841 SINGLE_DEPTH_NESTING) 842 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) 843 844 extern struct sock *sk_alloc(struct net *net, int family, 845 gfp_t priority, 846 struct proto *prot); 847 extern void sk_free(struct sock *sk); 848 extern void sk_release_kernel(struct sock *sk); 849 extern struct sock *sk_clone(const struct sock *sk, 850 const gfp_t priority); 851 852 extern struct sk_buff *sock_wmalloc(struct sock *sk, 853 unsigned long size, int force, 854 gfp_t priority); 855 extern struct sk_buff *sock_rmalloc(struct sock *sk, 856 unsigned long size, int force, 857 gfp_t priority); 858 extern void sock_wfree(struct sk_buff *skb); 859 extern void sock_rfree(struct sk_buff *skb); 860 861 extern int sock_setsockopt(struct socket *sock, int level, 862 int op, char __user *optval, 863 int optlen); 864 865 extern int sock_getsockopt(struct socket *sock, int level, 866 int op, char __user *optval, 867 int __user *optlen); 868 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, 869 unsigned long size, 870 int noblock, 871 int *errcode); 872 extern void *sock_kmalloc(struct sock *sk, int size, 873 gfp_t priority); 874 extern void sock_kfree_s(struct sock *sk, void *mem, int size); 875 extern void sk_send_sigurg(struct sock *sk); 876 877 /* 878 * Functions to fill in entries in struct proto_ops when a protocol 879 * does not implement a particular function. 880 */ 881 extern int sock_no_bind(struct socket *, 882 struct sockaddr *, int); 883 extern int sock_no_connect(struct socket *, 884 struct sockaddr *, int, int); 885 extern int sock_no_socketpair(struct socket *, 886 struct socket *); 887 extern int sock_no_accept(struct socket *, 888 struct socket *, int); 889 extern int sock_no_getname(struct socket *, 890 struct sockaddr *, int *, int); 891 extern unsigned int sock_no_poll(struct file *, struct socket *, 892 struct poll_table_struct *); 893 extern int sock_no_ioctl(struct socket *, unsigned int, 894 unsigned long); 895 extern int sock_no_listen(struct socket *, int); 896 extern int sock_no_shutdown(struct socket *, int); 897 extern int sock_no_getsockopt(struct socket *, int , int, 898 char __user *, int __user *); 899 extern int sock_no_setsockopt(struct socket *, int, int, 900 char __user *, int); 901 extern int sock_no_sendmsg(struct kiocb *, struct socket *, 902 struct msghdr *, size_t); 903 extern int sock_no_recvmsg(struct kiocb *, struct socket *, 904 struct msghdr *, size_t, int); 905 extern int sock_no_mmap(struct file *file, 906 struct socket *sock, 907 struct vm_area_struct *vma); 908 extern ssize_t sock_no_sendpage(struct socket *sock, 909 struct page *page, 910 int offset, size_t size, 911 int flags); 912 913 /* 914 * Functions to fill in entries in struct proto_ops when a protocol 915 * uses the inet style. 916 */ 917 extern int sock_common_getsockopt(struct socket *sock, int level, int optname, 918 char __user *optval, int __user *optlen); 919 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, 920 struct msghdr *msg, size_t size, int flags); 921 extern int sock_common_setsockopt(struct socket *sock, int level, int optname, 922 char __user *optval, int optlen); 923 extern int compat_sock_common_getsockopt(struct socket *sock, int level, 924 int optname, char __user *optval, int __user *optlen); 925 extern int compat_sock_common_setsockopt(struct socket *sock, int level, 926 int optname, char __user *optval, int optlen); 927 928 extern void sk_common_release(struct sock *sk); 929 930 /* 931 * Default socket callbacks and setup code 932 */ 933 934 /* Initialise core socket variables */ 935 extern void sock_init_data(struct socket *sock, struct sock *sk); 936 937 /** 938 * sk_filter_release: Release a socket filter 939 * @fp: filter to remove 940 * 941 * Remove a filter from a socket and release its resources. 942 */ 943 944 static inline void sk_filter_release(struct sk_filter *fp) 945 { 946 if (atomic_dec_and_test(&fp->refcnt)) 947 kfree(fp); 948 } 949 950 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) 951 { 952 unsigned int size = sk_filter_len(fp); 953 954 atomic_sub(size, &sk->sk_omem_alloc); 955 sk_filter_release(fp); 956 } 957 958 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) 959 { 960 atomic_inc(&fp->refcnt); 961 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); 962 } 963 964 /* 965 * Socket reference counting postulates. 966 * 967 * * Each user of socket SHOULD hold a reference count. 968 * * Each access point to socket (an hash table bucket, reference from a list, 969 * running timer, skb in flight MUST hold a reference count. 970 * * When reference count hits 0, it means it will never increase back. 971 * * When reference count hits 0, it means that no references from 972 * outside exist to this socket and current process on current CPU 973 * is last user and may/should destroy this socket. 974 * * sk_free is called from any context: process, BH, IRQ. When 975 * it is called, socket has no references from outside -> sk_free 976 * may release descendant resources allocated by the socket, but 977 * to the time when it is called, socket is NOT referenced by any 978 * hash tables, lists etc. 979 * * Packets, delivered from outside (from network or from another process) 980 * and enqueued on receive/error queues SHOULD NOT grab reference count, 981 * when they sit in queue. Otherwise, packets will leak to hole, when 982 * socket is looked up by one cpu and unhasing is made by another CPU. 983 * It is true for udp/raw, netlink (leak to receive and error queues), tcp 984 * (leak to backlog). Packet socket does all the processing inside 985 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets 986 * use separate SMP lock, so that they are prone too. 987 */ 988 989 /* Ungrab socket and destroy it, if it was the last reference. */ 990 static inline void sock_put(struct sock *sk) 991 { 992 if (atomic_dec_and_test(&sk->sk_refcnt)) 993 sk_free(sk); 994 } 995 996 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, 997 const int nested); 998 999 static inline void sk_set_socket(struct sock *sk, struct socket *sock) 1000 { 1001 sk->sk_socket = sock; 1002 } 1003 1004 /* Detach socket from process context. 1005 * Announce socket dead, detach it from wait queue and inode. 1006 * Note that parent inode held reference count on this struct sock, 1007 * we do not release it in this function, because protocol 1008 * probably wants some additional cleanups or even continuing 1009 * to work with this socket (TCP). 1010 */ 1011 static inline void sock_orphan(struct sock *sk) 1012 { 1013 write_lock_bh(&sk->sk_callback_lock); 1014 sock_set_flag(sk, SOCK_DEAD); 1015 sk_set_socket(sk, NULL); 1016 sk->sk_sleep = NULL; 1017 write_unlock_bh(&sk->sk_callback_lock); 1018 } 1019 1020 static inline void sock_graft(struct sock *sk, struct socket *parent) 1021 { 1022 write_lock_bh(&sk->sk_callback_lock); 1023 sk->sk_sleep = &parent->wait; 1024 parent->sk = sk; 1025 sk_set_socket(sk, parent); 1026 security_sock_graft(sk, parent); 1027 write_unlock_bh(&sk->sk_callback_lock); 1028 } 1029 1030 extern int sock_i_uid(struct sock *sk); 1031 extern unsigned long sock_i_ino(struct sock *sk); 1032 1033 static inline struct dst_entry * 1034 __sk_dst_get(struct sock *sk) 1035 { 1036 return sk->sk_dst_cache; 1037 } 1038 1039 static inline struct dst_entry * 1040 sk_dst_get(struct sock *sk) 1041 { 1042 struct dst_entry *dst; 1043 1044 read_lock(&sk->sk_dst_lock); 1045 dst = sk->sk_dst_cache; 1046 if (dst) 1047 dst_hold(dst); 1048 read_unlock(&sk->sk_dst_lock); 1049 return dst; 1050 } 1051 1052 static inline void 1053 __sk_dst_set(struct sock *sk, struct dst_entry *dst) 1054 { 1055 struct dst_entry *old_dst; 1056 1057 old_dst = sk->sk_dst_cache; 1058 sk->sk_dst_cache = dst; 1059 dst_release(old_dst); 1060 } 1061 1062 static inline void 1063 sk_dst_set(struct sock *sk, struct dst_entry *dst) 1064 { 1065 write_lock(&sk->sk_dst_lock); 1066 __sk_dst_set(sk, dst); 1067 write_unlock(&sk->sk_dst_lock); 1068 } 1069 1070 static inline void 1071 __sk_dst_reset(struct sock *sk) 1072 { 1073 struct dst_entry *old_dst; 1074 1075 old_dst = sk->sk_dst_cache; 1076 sk->sk_dst_cache = NULL; 1077 dst_release(old_dst); 1078 } 1079 1080 static inline void 1081 sk_dst_reset(struct sock *sk) 1082 { 1083 write_lock(&sk->sk_dst_lock); 1084 __sk_dst_reset(sk); 1085 write_unlock(&sk->sk_dst_lock); 1086 } 1087 1088 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); 1089 1090 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); 1091 1092 static inline int sk_can_gso(const struct sock *sk) 1093 { 1094 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); 1095 } 1096 1097 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); 1098 1099 static inline int skb_copy_to_page(struct sock *sk, char __user *from, 1100 struct sk_buff *skb, struct page *page, 1101 int off, int copy) 1102 { 1103 if (skb->ip_summed == CHECKSUM_NONE) { 1104 int err = 0; 1105 __wsum csum = csum_and_copy_from_user(from, 1106 page_address(page) + off, 1107 copy, 0, &err); 1108 if (err) 1109 return err; 1110 skb->csum = csum_block_add(skb->csum, csum, skb->len); 1111 } else if (copy_from_user(page_address(page) + off, from, copy)) 1112 return -EFAULT; 1113 1114 skb->len += copy; 1115 skb->data_len += copy; 1116 skb->truesize += copy; 1117 sk->sk_wmem_queued += copy; 1118 sk_mem_charge(sk, copy); 1119 return 0; 1120 } 1121 1122 /* 1123 * Queue a received datagram if it will fit. Stream and sequenced 1124 * protocols can't normally use this as they need to fit buffers in 1125 * and play with them. 1126 * 1127 * Inlined as it's very short and called for pretty much every 1128 * packet ever received. 1129 */ 1130 1131 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) 1132 { 1133 sock_hold(sk); 1134 skb->sk = sk; 1135 skb->destructor = sock_wfree; 1136 atomic_add(skb->truesize, &sk->sk_wmem_alloc); 1137 } 1138 1139 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) 1140 { 1141 skb->sk = sk; 1142 skb->destructor = sock_rfree; 1143 atomic_add(skb->truesize, &sk->sk_rmem_alloc); 1144 sk_mem_charge(sk, skb->truesize); 1145 } 1146 1147 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, 1148 unsigned long expires); 1149 1150 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); 1151 1152 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 1153 1154 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb) 1155 { 1156 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces 1157 number of warnings when compiling with -W --ANK 1158 */ 1159 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= 1160 (unsigned)sk->sk_rcvbuf) 1161 return -ENOMEM; 1162 skb_set_owner_r(skb, sk); 1163 skb_queue_tail(&sk->sk_error_queue, skb); 1164 if (!sock_flag(sk, SOCK_DEAD)) 1165 sk->sk_data_ready(sk, skb->len); 1166 return 0; 1167 } 1168 1169 /* 1170 * Recover an error report and clear atomically 1171 */ 1172 1173 static inline int sock_error(struct sock *sk) 1174 { 1175 int err; 1176 if (likely(!sk->sk_err)) 1177 return 0; 1178 err = xchg(&sk->sk_err, 0); 1179 return -err; 1180 } 1181 1182 static inline unsigned long sock_wspace(struct sock *sk) 1183 { 1184 int amt = 0; 1185 1186 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 1187 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); 1188 if (amt < 0) 1189 amt = 0; 1190 } 1191 return amt; 1192 } 1193 1194 static inline void sk_wake_async(struct sock *sk, int how, int band) 1195 { 1196 if (sk->sk_socket && sk->sk_socket->fasync_list) 1197 sock_wake_async(sk->sk_socket, how, band); 1198 } 1199 1200 #define SOCK_MIN_SNDBUF 2048 1201 #define SOCK_MIN_RCVBUF 256 1202 1203 static inline void sk_stream_moderate_sndbuf(struct sock *sk) 1204 { 1205 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { 1206 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); 1207 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); 1208 } 1209 } 1210 1211 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); 1212 1213 static inline struct page *sk_stream_alloc_page(struct sock *sk) 1214 { 1215 struct page *page = NULL; 1216 1217 page = alloc_pages(sk->sk_allocation, 0); 1218 if (!page) { 1219 sk->sk_prot->enter_memory_pressure(sk); 1220 sk_stream_moderate_sndbuf(sk); 1221 } 1222 return page; 1223 } 1224 1225 /* 1226 * Default write policy as shown to user space via poll/select/SIGIO 1227 */ 1228 static inline int sock_writeable(const struct sock *sk) 1229 { 1230 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); 1231 } 1232 1233 static inline gfp_t gfp_any(void) 1234 { 1235 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL; 1236 } 1237 1238 static inline long sock_rcvtimeo(const struct sock *sk, int noblock) 1239 { 1240 return noblock ? 0 : sk->sk_rcvtimeo; 1241 } 1242 1243 static inline long sock_sndtimeo(const struct sock *sk, int noblock) 1244 { 1245 return noblock ? 0 : sk->sk_sndtimeo; 1246 } 1247 1248 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) 1249 { 1250 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; 1251 } 1252 1253 /* Alas, with timeout socket operations are not restartable. 1254 * Compare this to poll(). 1255 */ 1256 static inline int sock_intr_errno(long timeo) 1257 { 1258 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; 1259 } 1260 1261 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 1262 struct sk_buff *skb); 1263 1264 static __inline__ void 1265 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) 1266 { 1267 ktime_t kt = skb->tstamp; 1268 1269 if (sock_flag(sk, SOCK_RCVTSTAMP)) 1270 __sock_recv_timestamp(msg, sk, skb); 1271 else 1272 sk->sk_stamp = kt; 1273 } 1274 1275 /** 1276 * sk_eat_skb - Release a skb if it is no longer needed 1277 * @sk: socket to eat this skb from 1278 * @skb: socket buffer to eat 1279 * @copied_early: flag indicating whether DMA operations copied this data early 1280 * 1281 * This routine must be called with interrupts disabled or with the socket 1282 * locked so that the sk_buff queue operation is ok. 1283 */ 1284 #ifdef CONFIG_NET_DMA 1285 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) 1286 { 1287 __skb_unlink(skb, &sk->sk_receive_queue); 1288 if (!copied_early) 1289 __kfree_skb(skb); 1290 else 1291 __skb_queue_tail(&sk->sk_async_wait_queue, skb); 1292 } 1293 #else 1294 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) 1295 { 1296 __skb_unlink(skb, &sk->sk_receive_queue); 1297 __kfree_skb(skb); 1298 } 1299 #endif 1300 1301 static inline 1302 struct net *sock_net(const struct sock *sk) 1303 { 1304 #ifdef CONFIG_NET_NS 1305 return sk->sk_net; 1306 #else 1307 return &init_net; 1308 #endif 1309 } 1310 1311 static inline 1312 void sock_net_set(struct sock *sk, struct net *net) 1313 { 1314 #ifdef CONFIG_NET_NS 1315 sk->sk_net = net; 1316 #endif 1317 } 1318 1319 /* 1320 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. 1321 * They should not hold a referrence to a namespace in order to allow 1322 * to stop it. 1323 * Sockets after sk_change_net should be released using sk_release_kernel 1324 */ 1325 static inline void sk_change_net(struct sock *sk, struct net *net) 1326 { 1327 put_net(sock_net(sk)); 1328 sock_net_set(sk, hold_net(net)); 1329 } 1330 1331 static inline struct sock *skb_steal_sock(struct sk_buff *skb) 1332 { 1333 if (unlikely(skb->sk)) { 1334 struct sock *sk = skb->sk; 1335 1336 skb->destructor = NULL; 1337 skb->sk = NULL; 1338 return sk; 1339 } 1340 return NULL; 1341 } 1342 1343 extern void sock_enable_timestamp(struct sock *sk); 1344 extern int sock_get_timestamp(struct sock *, struct timeval __user *); 1345 extern int sock_get_timestampns(struct sock *, struct timespec __user *); 1346 1347 /* 1348 * Enable debug/info messages 1349 */ 1350 extern int net_msg_warn; 1351 #define NETDEBUG(fmt, args...) \ 1352 do { if (net_msg_warn) printk(fmt,##args); } while (0) 1353 1354 #define LIMIT_NETDEBUG(fmt, args...) \ 1355 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) 1356 1357 extern __u32 sysctl_wmem_max; 1358 extern __u32 sysctl_rmem_max; 1359 1360 extern void sk_init(void); 1361 1362 extern int sysctl_optmem_max; 1363 1364 extern __u32 sysctl_wmem_default; 1365 extern __u32 sysctl_rmem_default; 1366 1367 #endif /* _SOCK_H */ 1368